US10534995B2 - Network of intelligent machines - Google Patents

Network of intelligent machines Download PDF

Info

Publication number
US10534995B2
US10534995B2 US13/843,784 US201313843784A US10534995B2 US 10534995 B2 US10534995 B2 US 10534995B2 US 201313843784 A US201313843784 A US 201313843784A US 10534995 B2 US10534995 B2 US 10534995B2
Authority
US
United States
Prior art keywords
parameters
processing unit
items
machines
machine
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active, expires
Application number
US13/843,784
Other languages
English (en)
Other versions
US20140279717A1 (en
Inventor
Alysia M. Sagi-Dolev
Alon Zweig
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Qylur Intelligent Systems Inc
Original Assignee
Qylur Intelligent Systems Inc
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Priority to US13/843,784 priority Critical patent/US10534995B2/en
Application filed by Qylur Intelligent Systems Inc filed Critical Qylur Intelligent Systems Inc
Priority to PCT/US2014/019134 priority patent/WO2014149510A2/en
Priority to EP22156222.6A priority patent/EP4044066A1/en
Priority to RU2015139090A priority patent/RU2669527C2/ru
Priority to CN201480021265.8A priority patent/CN105122277B/zh
Priority to EP14711367.4A priority patent/EP2973239A2/en
Priority to CA2903041A priority patent/CA2903041C/en
Priority to JP2016500474A priority patent/JP6528040B2/ja
Priority to BR112015023372-4A priority patent/BR112015023372B1/pt
Publication of US20140279717A1 publication Critical patent/US20140279717A1/en
Priority to IL241056A priority patent/IL241056B/en
Priority to HK16106098.0A priority patent/HK1218174A1/zh
Assigned to QYLUR INTELLIGENT SYSTEMS, INC. reassignment QYLUR INTELLIGENT SYSTEMS, INC. ASSIGNMENT OF ASSIGNORS INTEREST (SEE DOCUMENT FOR DETAILS). Assignors: ZWEIG, Alon, SAGI-DOLEV, ALYSIA M.
Priority to JP2019053265A priority patent/JP6790160B2/ja
Priority to US16/741,551 priority patent/US11544533B2/en
Application granted granted Critical
Publication of US10534995B2 publication Critical patent/US10534995B2/en
Priority to US18/091,654 priority patent/US20230153657A1/en
Active legal-status Critical Current
Adjusted expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/50—Context or environment of the image
    • G06V20/52—Surveillance or monitoring of activities, e.g. for recognising suspicious objects
    • G06N3/0454—
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N5/00—Computing arrangements using knowledge-based models
    • G06N5/04—Inference or reasoning models
    • G06N5/043—Distributed expert systems; Blackboards
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/217—Validation; Performance evaluation; Active pattern learning techniques
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
    • G06F18/24133—Distances to prototypes
    • G06F18/24143—Distances to neighbourhood prototypes, e.g. restricted Coulomb energy networks [RCEN]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
    • G06V10/776—Validation; Performance evaluation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/60—Type of objects
    • G06V20/68—Food, e.g. fruit or vegetables

Definitions

  • the present invention relates generally to a system for processing data obtained from a plurality of intelligent machines and particularly for machines that change their internal states based on input that is shared between machines.
  • Today, computerized machines are used to perform tasks in almost all aspects of life, such as handling purchases at store checkout stands, and taking and tracking orders at Internet shopping sites, packaging and sorting merchandise, keeping track of inventory in warehouses, tracking automobile registration data, medical screening for various conditions, and detecting the presence of certain objects or conditions.
  • there are many machines at different locations handling similar tasks For example, hospitals may have different campuses with a number of MRI machines in different parts of the campuses.
  • grocery store chains may have many stores and warehouses across a large geographical area, each store having a number of checkout registers.
  • farmers and orchards may each have their own facilities to automatically sort their produce, like sorting apples into high and low grade.
  • sorting machines are often based on the appearance of the product, like in the case where a video camera is used to identify bad fruits based on an automatic classifier.
  • An intelligent system that eliminates the inefficiency and redundancy and increases the accuracy by allowing machines to coordinate, communicate, and learn from each other is desired.
  • the invention is a self-updating apparatus configured to characterize items or conditions.
  • the apparatus includes a memory storing parameters for different items, wherein the parameters are useful for characterizing the items, and a processing module.
  • the processing module is configured to automatically select sources from which to receive data, modify the parameters based on the data that is received and to select recipients of modified parameters. The selection of the sources and recipients is based on comparison of parameters between the processing module and the sources and between the processing module and the recipients, respectively.
  • the invention is a self-updating network of machines that are configured to characterize items or conditions.
  • the network of machines includes a plurality of machines taking measurements of items and a central processing unit.
  • Each of the machines has a processing unit and a measurement unit that collaborate to collect physical measurements from the items, generate parameters for characterizing the items based on the physical measurements, and select a subgroup of parameters to be transmitted to a particular recipient.
  • the central processing unit is configured to receive subgroups of parameters from the machines, update central parameters based on received subgroups of parameters, and send the updated central parameters to some of the plurality of machines.
  • the invention is a computer-implemented method of characterizing an item or condition.
  • the method entails obtaining measurements from an item; generating parameters for the item based on the measurements, wherein the parameters are useful for characterizing the item; comparing the parameters with new parameters received from a processing unit; selectively receiving at least some of the new parameters from the processing unit based on the comparison; and automatically modifying the parameters based on the new parameters.
  • FIG. 1 depicts a machine network system that includes a plurality of machines that communicate with each other and with a central processing unit.
  • FIG. 2 is a detailed depiction of the machines and the central processing unit.
  • FIG. 3 is a flowchart illustrating the parameter updating process.
  • FIG. 4 depicts an example embodiment where each machine is a sorting machine.
  • FIG. 5 depicts one of the machines of FIG. 4 in more detail.
  • FIG. 6 depicts an optical unit portion of the machine in FIG. 5 .
  • FIG. 7A is a computer image of fruit showing soft puff and crease as detected by the machine of FIG. 5 .
  • FIG. 7B is a histogram of the fruit surface corresponding to the image of FIG. 7A .
  • FIG. 8 is a histogram obtained from a surface of a fruit having sour rot.
  • FIG. 9 is a histogram obtained from a surface of a fruit having clear rot.
  • FIG. 10 is a histogram obtained from a surface of a fruit having a pebbled peel.
  • FIG. 11 is a histogram obtained from a surface of a fruit showing soft puff and crease condition.
  • FIG. 12 is a histogram obtained from a surface of a fruit showing a ridge and valley defect.
  • FIG. 13 is a histogram obtained from a fruit having a split or cut in the peel.
  • FIG. 14 is a histogram obtained from a fruit having clear puff and crease condition.
  • Embodiments are described herein in the context of machines that classify fruits according to their grades.
  • the embodiments provided herein are just examples and the scope of the invention is not limited to the applications or the embodiments disclosed herein.
  • the system of the invention may be useful for any type of equipment that is capable of automatically learning rules from examples (machine learning algorithms), including but not limited to a machine that employs artificial neural network and is capable of iterative learning, such as medical diagnostic machines, fault testing machines, and object identification machines.
  • “remotely located” means located in different forums, companies, organizations, institutions, and/or physical locations. Machines that are located on different floors of the same building, for example, could be remotely located from each other if the different floors host different organizations.
  • a “processing unit,” as used herein, includes both a central processing unit ( 20 ) and a machine ( 30 ) or a group of machines ( 50 ).
  • “Parameters,” as used herein, include central parameters and internal parameters.
  • the system of the disclosure is useful for coordinating information exchange among a plurality of machines.
  • This disclosure discusses a network of machines in communication with each other, which examines the collective body of data from the different machines to generate and modify a set of central parameters.
  • the machines may be remotely-located from the central processing unit and in different places around the world.
  • the different machines can learn from one another and utilize the “knowledge” gained from different machines to teach and improve its counter parts. For example, where the machines are fruit-sorting machines, the machines may learn and adjust to a trend that a new condition affecting citrus fruits is showing up at different locations around the world.
  • the central processing unit may be able to either figure out a way to detect this condition based on this data, or utilize the adjusted updated local central parameters in that machine, determine which geographical locations are susceptible to this condition, and transmit information and new central parameters to the machines in these locations that will help detect this new condition so the fruits with the new defect can be rejected.
  • the central processing unit sees and analyzes the data from a high level using a global network of detection machines.
  • the system of the invention allows an intelligent, better-informed understanding of a situation that cannot be provided by individual machines alone.
  • FIG. 1 depicts a machine network 10 that includes a central processing unit 20 in communication with a plurality of machines 30 via a network.
  • the central processing unit 20 is configured to receive and selectively transmit information to the machines 30 .
  • Each machine 30 may be one machine unit or a group of units, and typically includes hardware components for receiving items to be tested.
  • a plurality of machines 30 are grouped to form a “group” or “family” 50 of machines 30 that directly share data among themselves without going through the central processing unit 20 .
  • Machines 30 in a family 50 of machines often have a commonality, such as presence in a same general geographic region, configuration to handle specific fruit types (e.g., citrus fruit), or representing the same company.
  • the machines 30 test each item that is received and characterizes the item according to a set of internal parameters. For example, in the case of fruit-sorting machines, the machine 30 may characterize each fruit as “reject,” “juice,” “Grade B,” and “Grade A.”
  • the machines 30 are configured to transmit data to the central processing unit 20 , which collects data from all the machines 30 and develops its own central parameters.
  • the central processing unit 20 initially receives data from the machines 30 to self-train and generate its own set of central parameters. As more data is received, the central processing unit 20 refines and modifies its central parameters such that accuracy and breadth of the characterization is enhanced over time.
  • the machine 30 would typically be used to characterize an item, for example by detecting the presence of a condition.
  • the machine 30 may be a fruit sorting machine, a detection machine, a store checkout machine, a medical diagnostic machine, a fault detection machine etc.
  • the invention is not limited to being used with any particular type of machine.
  • the machine 30 could be part of a security check system at the entrance of a high-tech manufacturing facility, in which case it could be used to detect the presence of any digital storage devices that may be used for misappropriating intellectual property or technical data.
  • a machine at the entrance/exits of stores could be used to detect stolen merchandise.
  • a fault detection machine could detect micro cracks in air plane wings, and a medical diagnostic device could detect types of cancer, fractures or other conditions.
  • a fruit sorting machine would detect bruises or damage on the fruit. If the presence of a target item is detected, an alarm will be generated to invite an operator who can confirm the presence of the target item/condition, or to activate an automatic response such as locking the undesired object, re-directing it to a trash bin, opening a repair ticket, or placing a comment in a medical file.
  • each machine 30 has a unique set of strengths and weaknesses.
  • Each machine 30 sends data to other machines 30 and the central processing unit 20 and receives information from other machines 30 and the central processing unit 20 .
  • the communication between different machines as well as between the machines 30 and the central processing unit 20 can be achieved through any secure network using a predetermined protocol.
  • a processing unit determines which data should be sent to which other processing units based on data comparison among the machines 30 . For example, if data comparison reveals that Machine X has encountered items that Machine Y has not encountered yet, Machine X may transmit parameters for the items that Machine Y has not encountered to Machine Y, so that Machine Y will recognize the item when it first encounters the item. In another example, where a fig sorting machine 30 and an orange sorting machine 30 compare data with each other and other machines 30 , the fig sorting machine and the orange sorting machine may notice that some of the other machines sort objects that are generally round and share similar characteristics as figs and oranges.
  • a security check machine in Building A may frequently encounter USB devices carried by employees.
  • a security check machine in Building B may not have encountered USB devices from its employees and customers.
  • the machine at Building A may send parameters for USB devices to the machine at Building B. If a third machine at Building C already has its own parameters for USB devices, machines at Buildings A and C may compare their internal parameters and make any updates to further refine the parameters.
  • the machines and processing units can “learn” from each other by comparing and updating their parameters.
  • parameters that are missing in one processing unit are added by being received from another processing unit.
  • parameters that are different yet identify the same item triggers the processing units to modify one or more sets of parameters to strengthen the characterization capability.
  • each machine 30 or a group of machines 30 incorporates an artificial intelligence program and is able to learn or change their internal states based on input.
  • the machines 30 may learn about ordinary items as more items pass through it.
  • the machines may, for example, incorporate a neural network.
  • the synaptic weights and thresholds of the neural network are initialized and a first set of items are introduced to the machines to receive an output.
  • the output would be a classification assigned to each fruit (e.g., Grade A, Grade B, juice, reject).
  • a machine trainer will initially feed a randomly mixed batch of fruits to the machine 30 and provide input as to how each fruit is categorized, thereby “training” the machine 30 .
  • the machine 30 by using the measurements and the outcomes that each set of measurements was supposed to produce, generates a set of conditions for identifying how a fruit should be categorized.
  • the machine runs tests on the items, makes measurements, and generates a set of parameters for each item.
  • Each machine has a storage unit that records the parameters of all the items it encountered.
  • After the initial training with a set of fruits, each machine has a set of internal parameters that it uses to characterize the next fruit that is received. The more fruits a machine 30 or a group of machines 30 has seen, the more data points it will have in its memory and the more accurate the next characterization will be.
  • the internal parameters are continually modified to enhance the accuracy of characterization.
  • each machine 30 transmits the parameters of all the items it encountered to the central processing unit 20 .
  • the central processing unit 20 maintains a set of central parameters.
  • the central processing unit 20 processes the data received from the plurality of machines 30 in the system by running each input to generate and modify the central parameters, which are used to characterize the next fruit.
  • the central processing unit 20 also incorporates an artificial intelligence program. As the central processing unit 20 receives data from all the machines 30 in the network, it will develop a broader set of parameters that cover all the global possibilities. Furthermore, the central processing unit 20 will be able to analyze regional trends, unlike the machines 30 . Based on the trends and patterns it sees, the central processing unit 20 can prepare certain machines for new parameters they will encounter. Alternatively, machines 30 can directly share with each other data that they encountered, effectively “educating” one another.
  • the central processing unit 20 also receives external data 40 , such as intelligence data or any data to be selectively distributed to the machines 30 .
  • the external data 40 may include intelligence information or details about situations in certain regions. For example, suppose a situation where the machines are detection machines. If a valuable painting is stolen in Florence, Italy, the outside data can be used to inform the central processing unit 20 of this situation.
  • the central processing unit 20 can, in response, adjust the parameters to heighten the sensitivity for paintings and transmit the adjusted parameters to the machines so that the machines will almost immediately be “looking for” the stolen painting. Likewise, if a stadium has long lines that are moving slowly, a request can be input to lower the sensitivity level of the machines at the entrance to help move the lines along faster.
  • information regarding expected weather trends in a given geography can alert the system to heighten the detection of certain types of damage that are correlated to this weather.
  • a turbine safety inspection machine may learn pattern of certain blade damage due to increased feather residues from increased bird migration during certain seasons and geographies and adjust those machines to increase sensitivity for those inspection machines each year at that period and in that region.
  • the external data 50 may be input by the machine trainer 40 or from another source.
  • FIG. 2 depicts one embodiment of the machine 30 and the central processing unit 20 .
  • Each machine 30 has a processing module 32 that employs artificial intelligence and a memory 38 that stores internal parameters.
  • the processing module 32 and the memory 38 are together referred to as a “processing unit” ( 32 + 38 ).
  • the machine 30 is configured to receive items, move the received items, for example with a moving mechanism, and subject each item to one or more tests via a test module 34 .
  • the test may be a simple determination of shape or weight, and may be a more complex form of imaging, as well as any other known tests that would help analyze or detect a condition.
  • a characterization module 36 characterizes the item.
  • the test module 34 and the characterization module 36 are together referred to as a “measurement unit” ( 34 + 36 ), and includes physical components for receiving, holding, and testing items. If more information is needed to characterize the item, the machine 30 requests extra information from an external source, such as an operator or additional sensors. In characterizing the item, the machine 30 uses the internal parameters stored in a memory 38 . The internal parameters were previously generated by correlating items with different conditions with their characterization. Hence, the characterization includes comparison of the measurements against the internal parameters. As more extra information is received, each machine may update or modify its set of internal parameters.
  • the machine 30 has a receiver/transmitter 39 for exchanging information via the network.
  • the central processing unit 20 includes a processing module 24 that includes an artificial intelligence program, and a memory 22 that includes a machine database for storing central parameters and data about the different machines in the network.
  • the processing module 24 and the memory 22 are together referred to as the “processing unit” ( 24 + 22 ).
  • the central processing unit 20 generates its own set of central parameters based on the measurement and characterization data it received from the machines 30 .
  • the central parameters are likely to be much more extensive and inclusive compared to the local internal parameters on any single machine 30 because while each machine 30 only encounters the items that it directly handles, the central processing unit 20 has a global perspective.
  • the machine database keeps track of all the machines that send information to it.
  • the central processing unit 20 may tag the data with a machine ID to track which machine, group of machines, or family of machines that share knowledge, the data came from. This way, the central processing unit 20 can catch any trends such as weather or other external common phenomena, or be on the lookout for a pattern that may be a warning sign.
  • the central processing unit 20 also uses the machine database to determine which machine will benefit from a new update/modification to the parameters.
  • the central processing unit 20 and each machine 30 has a receiving portion 26 and a transmitting portion 28 for communicating to other machines 30 and processing units in the network.
  • the receiving portion 26 and the transmitting portion 28 may be one physical component.
  • the central processing unit 20 also receives external data 40 from a source other than the machines 30 .
  • the processing module 32 of a machine 30 determines that there is an unusual situation at hand or the situation may need a warning, it generates an alert via the alert generator.
  • the processing module 32 of a machine 30 determines that there is an unusual situation at hand or the situation may need a warning, it generates an alert via the alert generator.
  • the processing module 32 of a machine 30 determines that there is an unusual situation at hand or the situation may need a warning, it generates an alert via the alert generator.
  • the processing module 32 of a machine 30 determines that there is an unusual situation at hand or the situation may need a warning, it generates an alert via the alert generator.
  • the processing module 32 of a machine 30 determines that there is an
  • both the machines 30 and the central processing unit 20 can include a user interface for communicating with an operator and/or machine trainer.
  • FIG. 3 illustrates the iterative parameter updating process 70 of the machine network system 10 , as well as the iterative data flow between the machines 30 and the central processing unit 20 .
  • the iterative data flow may happen directly between different machines 30 , or between machine groups (each “machine group” includes a plurality of machines).
  • a machine 30 receives items and subjects each item to a test to obtain measurements (step 71 ).
  • measurements are then compared against these parameters to determine an outcome (step 73 ).
  • the machine 30 concludes that no new item/situation is encountered and proceeds to characterize or process the item consistently, the way that it is trained to process items with those parameters (step 77 ).
  • the machine 30 may store, either in a local storage unit or at the central processing unit 20 , data from the scan regardless of the outcome. If the measurements do not match any previously encountered set of parameters well enough (step 75 —“yes”), an alert is generated to either trigger an automated response or alert an operator (step 79 ). The operator examines the item, reviews the measurements, and subjects the item to additional tests if desired to come up with a characterization. In some embodiments, the machine 30 collects additional information.
  • the operator then provides feedback (extra information) to the machine by inputting his characterization (step 81 ).
  • the machine updates its parameters to incorporate the extra information that it just received (step 83 ).
  • the measurements that triggered the alert and the operator input are either retained within the machine 30 , and or transmitted to other machines 30 , and or transmitted to the central processing unit 20 (step 85 ).
  • the machines 30 , groups of machines 50 , or the central processing unit 20 receives measurements and characterizations from machines in the machine network 10 , which are often at different locations (step 91 ).
  • the central processing unit 20 , the machines 30 , and/or groups of machines 50 receive data independently of the machines 30 and has its set of central parameters (step 93 ) from previous training.
  • the training may be based on its own internal data sets or data sets received from other machines 30 , families of machines 50 and/or the central processing unit 20 .
  • the central processing unit 20 also receives external data (step 95 ).
  • the machines 30 and/or the central processing unit 20 continually modifies its central parameters 94 based on the measurement data it receives from the machines 30 , families of machines 50 , and the central processing unit 20 and the extra information 81 that pertains to previously un-encountered items/conditions.
  • Central parameters help the machines 30 , family of machines 50 , and the central processing units 20 to identify which items are being encountered by almost all the machines, so the parameters for that items can be strengthened.
  • Central parameters may be used to selectively increase the “resolution” of detection.
  • the central processing unit 20 , the machines 30 and/or families of machines 50 identifies, either for its self or other, machines that would benefit from the updated parameters, e.g.
  • the machines that would encounter the condition/item that is affected by the updated parameter (step 97 ). For example, where the newly found condition is due to a fruit disease that is only in a certain region, parameters for the condition would not be sent to machines at other locations. On the other hand, if the parameters pertain to a condition that is applicable to any location (e.g., a bruise) the parameters may be sent to all the machines 30 . The updated parameters are then transmitted to the identified select machines (step 99 ).
  • the machines 30 that receive the modified central parameters may further modify their own internal parameters to reflect the modified central parameters. This way, the machines 30 , other machines 30 , other families of machines 50 and the central processing unit 20 are continually teaching and learning from one another.
  • the machine 30 in the machine network 10 learns from multiple sources: 1 ) items that pass through the machine, 2 ) extra information received, e.g. from the local operator, 3 ) updated parameters received from the central processing unit 20 , 4 ) new data and updated parameters from itself, 5 ) data and updated parameters from other machines 30 , or families of machines 50 .
  • FIG. 4 depicts an embodiment where each machine 30 is a sorting machine 100 .
  • the sorting machine 100 may be a fruit sorting machine that incorporates the features described in U.S. Pat. No. 5,845,002 and is enhanced with the processing unit 32 + 38 to include artificial intelligence and internal parameters.
  • the invention is not limited to being used with any particular type of machine, the disclosure is provided to provide a concrete example of how the processing module 32 in machine 30 and or central processing unit 20 may be used.
  • FIG. 5 shows select parts of the sorting machines 100 in more detail.
  • the sorting machine 100 includes a conventional conveyor line 112 upon which a plurality of items 114 are conveyed. For simplicity of illustration, this particular depiction does not show the processing unit, although the processing unit is part of the sorting machine 100 .
  • the items 114 are fruit (e.g., citrus fruit) in the context of this disclosure, although this is not a limitation of the invention.
  • the particular sorting machine 100 may be suitable for items that are generally spherical and have a topographic surface texture. In other embodiments, the sorting machines 100 may be replaced by medical examination machines, object screening machines, etc.
  • the conveyor 112 transports the fruit 114 into an optical housing 116 , where the fruit is illuminated at an inspection station 118 within an optical housing 116 .
  • the conveyor 112 transports and orients the fruit 114 to control the presentation of the fruit 114 for imaging.
  • the conveyor is designed to provide a maximum optical exposure of fruit 114 at inspection station 118 .
  • Conveyor system 112 in the illustrated embodiment includes driven spools to rotate the fruit 114 .
  • the fruit 114 is rotated in a retrograde direction as it moves through the inspection station 118 to at least partially compensate for its forward motion down conveyor 112 .
  • the fruit 114 is rotated so that the same surface tends to remain facing a camera 130 during an extended time exposure to allow complete and reliable imaging. This may, of course, be time-synchronized by means well known in the art.
  • the fruit 14 is illuminated by a pair of light sources 122 , 124 .
  • the light sources 122 , 124 are focused on the fruit 114 from below and may further be provided with conventional optics to assist in providing optimal illumination of the surface of the fruit 114 .
  • the optical sources 22 , 24 may be optical fibers, or laser beams or light beams formed by LEDs. Alternatively, a single light source may be utilized and may be optically divided into two optical sources 22 , 24 .
  • the light sources 22 , 24 (or a single light source) provide the incident light that will be scattered within the fruit to cause it to glow.
  • the frequency or frequency spectrum of the light is selected based on the optical properties of the object to be inspected, to produce the desired scattering within the object, and the resultant projection of that glow through the surface thereof. With citrus fruit, the ordinary visible spectrum may suffice.
  • the camera 130 is coupled to a texture mode computer 134 .
  • the texture mode computer 134 is a personal computer coupled to both a master computer 136 which runs the functions of the conveyor and sorting systems and to an input/output computer 138 , which provides user input and output access to the system 100 .
  • the texture analysis of the fruit 114 is made by the texture mode computer 134 .
  • input through input/output computer 138 to master remote computer 136 will implement a sorting operation as dictated by texture mode computer 134 at a plurality of sorting stations 140 , which may include solenoid-actuated ejection fingers upon which the fruit 114 rides, and by which the fruit 114 is ejected from the conveyor line 112 into appropriate sorting bins 142 or secondary conveyors.
  • the texture module of the sorting machine 100 is made up of three subsystems that include the lighting and optics (including the optical housing 116 ), imaging as provided by the cameras 30 and mirrors 126 a , 126 b , 128 a , 128 b , and image processing within the texture mode computer 134 .
  • the central input/output computer 138 and the master remote computer 136 are conventional and are substantially the same as used in prior art classification and sorting apparatus.
  • the central input/output computer 138 provides for system control including providing for all aspects of user interface, selection for input and output of various classification parameters, and for determining conveyor paths in the machine 100 where multiple lanes for the conveyor 112 are provided in a more complex array than the simple linear depiction of FIG. 4 .
  • the inspection station 118 is appropriately baffled as desired, either to provide flat black nonreflecting surface to avoid spurious images, or to include reflective surfaces if desired to increase the light intensity incident upon the fruit.
  • the glow from light scattered within the fruit 114 and projected through its peel is reflected from lower mirrors 126 a , 126 b , and from there to upper mirrors 128 a , 128 b .
  • a CCD matrix or scanning camera 130 has its optics 132 focused on the upper mirrors 128 a , 128 b to capture, in a single computer image, virtually the entire exterior surface of a hemisphere of fruit 114 .
  • the first hemispheric image of the fruit 114 is reflected by the lower right mirror 127 a to the upper left mirror 129 a and from there to the first camera 130 a .
  • the image of that first hemisphere is also reflected by the lower left mirror 127 b into upper right mirror 129 b in the first camera 130 a.
  • the image of the second hemisphere of the fruit 114 is reflected by the lower right mirror 127 c to the upper left mirror 129 c , and from the lower left mirror 127 d to the upper left mirror 129 d , both resultant images being reflected into the other camera 130 b.
  • the lighting system uses two tungsten Halogen projection lamps 122 , 124 situated on opposite sides of the fruit 114 and below the fruit centerline.
  • the lamps emit enough light of the proper frequency or spectrum incident on the fruit 114 to create a glowing effect transmitted through the peel/skin of the fruit 114 that can be detected by a camera.
  • the fruit will provide a glowing effect to the camera provided that the positioning, intensity, and frequency/spectrum of the light is such that the light penetration into the peel or rind of fruit 114 occurs and is scattered therewithin to provide a glowing effect through the peel.
  • time exposure of the imaging is electronically controlled.
  • Electronic control of the time exposure compensates for any difference in the intensity of the glow due to differences in fruit size and peel thickness. This can be determined during the initial part of the run and appropriate corrections, either automatic or manual, may be entered through the input/output controller 138 .
  • Automatic control may be effected by user of a photodiode 144 mounted on each camera 130 to generate an output frequency, by a frequency generator (not shown), which depends upon the amount of light sensed by each photodiode.
  • a frequency generator (not shown)
  • the exposure time on the CCD chip within cameras 30 is controlled.
  • the texture mode computer 134 performs image processing and passes the classification information to the rest of the system for final drop-out selection according to means known in the art.
  • the first step is to drop out invalid information such as reflected light intensities from the light sources 122 , 124 which do not constitute the glow from light scattered within the fruit 114 and emerging through its peel.
  • FIG. 7A bright portions 146 of an actual computer image of a smooth fruit peel are depicted. Two images of fruit 114 are shown in FIG. 7A , depicting in essence the two hemispherical views of the fruit.
  • regions 146 of the graphic image because of their distinctively higher intensity levels, can be eliminated as portions of the graphic information signal carrying no information about the topographic surface texture.
  • a scan of the fruit surface is made to provide maximum, minimum, and standard deviation of the intensity of the entire pixel pattern constituting the image, to provide an indication if there are intensity variations in the image which could constitute surface defects requiring further examination, such as puff and crease, peel, cuts, punctures, etc.
  • a puff in a citrus fruit is an area of the peel which is slightly detached from the underlying meat, and thus will be slightly swollen or “puffed out.”
  • a crease is the reverse, in which a portion of the rind surface has been depressed relative to adjoining areas.
  • the graphic image is checked for high frequency data which, for example, would be indicative of pebbliness of the fruit surface.
  • the data derived from the fruit 114 can then be fed back to the master computer 136 for classification purposes according to predefined criteria.
  • the type of defect can then be determined by applying a series of data filters to identify them.
  • the high pass data filter can be used to search for cuts or punctures.
  • a low pass filter with blob analysis, tracing and aspect ratio of areas of greater intensity is useful to identify puff and crease and to distinguish it from rot.
  • a series of checks to show peak intensities over standard deviation values can be used to identify the degree of defect within a category of defect, such as puff and crease.
  • the size of the fruit as a whole is compared with the area affected in order to generate a percentage value for the defect of the affected surface.
  • Other defects, such as rot or breaks in the rind may not be subject to a percentage evaluation, but may constitute a cause for immediate rejection of the fruit regardless of the percentage of the affected area of the fruit.
  • FIG. 7A in which a computer image of a smooth orange rind is depicted, illustrates the double image from the reflected image provided to the camera.
  • Brightened areas 146 from the illumination source are eliminated as not containing information relevant to the nature of the peel condition.
  • Statistical information is then taken of the entire graphic image to obtain maxima, minima, and standard deviations to characterize the intensity variations of the image pixels. In this case, the statistical deviations which would be returned would indicate that the fruit was smooth and well within the acceptable range.
  • further statistical analysis would not be performed, and the fruit position tagged within the sorting machine 100 and carried down conveyor 112 to be routed to the appropriate sorting bin 142 or secondary conveyor, or for analysis and classification according to additional methods and criteria.
  • a typical scan line 148 is taken across one portion of the two hemispherical images in FIG. 7A .
  • Scan line intensity is then depicted in the histogram of FIG. 7B where intensity is graphed against the vertical scale and positioned along the scan line along the horizontal scale with end 150 corresponding to the left end of the histogram of FIG. 7B and end 152 of scan line 148 corresponding to the right end of the histogram of FIG. 7B .
  • a visual examination of the histogram of FIG. 7B indicates variations of pixel intensity maintained within a range of values with a fairly limited deviation from a mean, to provide a pattern quite different from the histograms depicted in FIGS. 8-14 , wherein various fruit defects are illustrated.
  • the histograms can be characterized by meaningful statistical parameters, and through those parameters, sorted into categories to reliably identify the topographic surface texture of the fruit 114 .
  • FIG. 8 depicts an intensity histogram taken from a computer image of a fruit 114 that is blemished by a surface decomposition known as sour rot. As compared to the histogram of FIG. 7B , there is a much wider variation between maxima and minima, and deviation from the mean is much greater than in the case of FIG. 7B .
  • FIG. 9 depicts an intensity histogram taken from a computer image of a fruit 114 that is characterized by a clear rot skin blemish. The histogram shows a large peak sharply falling off to an average pixel intensity.
  • FIG. 9 depicts an intensity histogram taken from a computer image of a fruit 114 that is characterized by a clear rot skin blemish. The histogram shows a large peak sharply falling off to an average pixel intensity.
  • FIG. 10 depicts an intensity histogram taken from a computer image of a fruit 114 that is characterized by high porosity or a pebbled surface that some consumers may dislike.
  • FIG. 11 depicts an intensity histogram taken from a computer image of a fruit 114 whose surface is blemished by a condition known as soft puff and crease
  • FIG. 12 is a histogram taken from a fruit 114 whose surface is blemished by a defect known as ridge and valley.
  • FIG. 13 depicts a histogram taken from a fruit 114 with “fracture,” which include splits, cuts, punctures, and scrapes.
  • FIG. 14 depicts a histogram taken from a fruit 114 with a skin defect called clear puff and crease.
  • Each of the histograms described above may be saved in the memory of the machines as part of predefined conditions for characterizing the fruit.
  • the machine 30 Upon taking a measurement from a newly received fruit, the machine 30 will subject it to the imaging test to produce an image similar to that disclosed in FIG. 7A and generate a histogram. The histogram will be then compared against the stored histograms that indicate certain conditions/defects to make a characterization.
  • Sorting machines 100 may be placed in different facilities such as farms and orchards, possibly in different parts of the world. Each sorting machine 100 would initially be “trained” by its sets of data obtained from samples or an operator who runs a representative sample of fruits through the machine and provides input as to in which bin 142 each fruit in the sample should be placed. The sorting machine 100 develops its own set of parameters based on the sample fruits and the inputs, and uses these parameters to categorize the next fruit it encounters. More specifically, in step 71 , the fruit is imaged as two hemispheres, in the manner shown in FIG.
  • step 73 the machine compares the histogram of the current fruit against its internal parameters. If there is a substantially close match between the histogram of the current fruit and one of the previously-generated histograms (step 75 —“no”), the current fruit will be sorted or categorized into the same bin 142 as the previous fruit that generated the similar histogram (step 77 ).
  • step 79 an alert is generated (step 79 ).
  • the data is used to train and determine how the fruit should be categorized, and tells the machine 100 how it should be categorized (step 81 ).
  • the machine 100 modifies or updates its internal parameters with this new data (step 83 ) and categorizes the current fruit according to the operator input (step 77 ).
  • the machines 30 , family of machines 50 , and/or central processing units 20 initially receive the scan data and categorization data from selections of machines, and in some cases all the machines in the network (step 91 ), and generates its own central parameters (step 93 ).
  • the central parameters may not be exactly the same as the local parameters on any one machine 100 since the machines 30 and or processing units 20 “see” more fruits than any single machine 100 in the network, and is presumably exposed to many more variations and conditions than any single machine 100 .
  • the central parameters thus, may be broader in the range of defects that are covered and able to distinguish defects with a higher resolution.
  • the machine 30 , family of machines 50 , and/or the central processing units 20 also receives any external data (step 95 ).
  • the external data might be a request from a Department of Agriculture to report all cases of a specific condition, or local weather conditions.
  • the machine 30 , family of machines 50 , and/or central processing units 20 then identifies the machines 100 that they should receive data from and that should receive the updated/modified central parameters (step 97 ). For example, if the update/modification to the parameters pertains to a pebbliness of the fruit skin, this update would be sent to machines 100 whose primary function is to sort fruits to send to various grocery stores. However, the modified parameters would not be sent to machines at a juicing factory because the texture of the fruit skin would not matter much to the juicing process, which typically happens after the skin is removed. At the same time, the machine 30 , family of machines 50 , and/or central processing units 20 also determines that all the machines 100 in the network should receive the request from the external data. The data and or parameters are then transmitted to the selected machines (step 99 ).
  • the processing units may be implemented with or involve one or more computer systems.
  • the computer system is not intended to suggest any limitation as to the scope of use or functionality of described embodiments.
  • the computer system includes at least one processor and memory.
  • the processor executes computer-executable instructions and may be a real or a virtual processor.
  • the computer system may include a multi-processing system which includes multiple processing units for executing computer-executable instructions to increase processing power.
  • the memory may be volatile memory (e.g., registers, cache, random access memory (RAM)), non-volatile memory (e.g., read only memory (ROM), electrically erasable programmable read only memory (EEPROM), flash memory, etc.), or combination thereof.
  • the memory may store software for implementing various embodiments of the disclosed concept.
  • the computing device may include components such as memory/storage, one or more input devices, one or more output devices, and one or more communication connections.
  • the storage may be removable or non-removable, and includes magnetic disks, magnetic tapes or cassettes, compact disc-read only memories (CD-ROMs), compact disc rewritables (CD-RWs), digital video discs (DVDs), or any other medium which may be used to store information and which may be accessed within the computing device.
  • the storage may store instructions for the software implementing various embodiments of the present invention.
  • the input device(s) may be a touch input device such as a keyboard, mouse, pen, trackball, touch screen, or game controller, a voice input computing device, a scanning computing device, a digital camera, or another device that provides input to the computing device.
  • the output computing device(s) may be a display, printer, speaker, or another computing device that provides output from the computing device.
  • the communication connection(s) enable communication over a communication medium to another computing device or system.
  • the communication medium conveys information such as computer-executable instructions, audio or video information, or other data in a modulated data signal.
  • a modulated data signal is a signal that has one or more of its characteristics set or changed in such a manner as to encode information in the signal.
  • communication media includes wired or wireless techniques implemented with an electrical, optical, RF, infrared, acoustic, or other carrier.
  • an interconnection mechanism such as a bus, controller, or network may interconnect the various components of the computer system.
  • operating system software may provide an operating environment for software's executing in the computer system, and may coordinate activities of the components of the computer system.
  • Computer-readable media are any available media that may be accessed within a computer system.
  • Computer-readable media include memory, storage, communication media, and combinations thereof.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • Data Mining & Analysis (AREA)
  • Computing Systems (AREA)
  • Software Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • General Health & Medical Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Multimedia (AREA)
  • Computational Linguistics (AREA)
  • Mathematical Physics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Databases & Information Systems (AREA)
  • Medical Informatics (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Molecular Biology (AREA)
  • Biophysics (AREA)
  • Biomedical Technology (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Evolutionary Biology (AREA)
  • Investigating Materials By The Use Of Optical Means Adapted For Particular Applications (AREA)
  • Image Processing (AREA)
  • Arrangements For Transmission Of Measured Signals (AREA)
  • Image Analysis (AREA)
  • Facsimile Image Signal Circuits (AREA)
  • Computer And Data Communications (AREA)
US13/843,784 2013-03-15 2013-03-15 Network of intelligent machines Active 2034-05-06 US10534995B2 (en)

Priority Applications (14)

Application Number Priority Date Filing Date Title
US13/843,784 US10534995B2 (en) 2013-03-15 2013-03-15 Network of intelligent machines
EP22156222.6A EP4044066A1 (en) 2013-03-15 2014-02-27 Network of intelligent machines
RU2015139090A RU2669527C2 (ru) 2013-03-15 2014-02-27 Сеть интеллектуальных машин
CN201480021265.8A CN105122277B (zh) 2013-03-15 2014-02-27 智能机器的网络
EP14711367.4A EP2973239A2 (en) 2013-03-15 2014-02-27 Network of intelligent machines
CA2903041A CA2903041C (en) 2013-03-15 2014-02-27 Network of intelligent machines
JP2016500474A JP6528040B2 (ja) 2013-03-15 2014-02-27 インテリジェントマシンのネットワーク
BR112015023372-4A BR112015023372B1 (pt) 2013-03-15 2014-02-27 Aparelho de auto-atualização e método implementado por computador
PCT/US2014/019134 WO2014149510A2 (en) 2013-03-15 2014-02-27 Network of intelligent machines
IL241056A IL241056B (en) 2013-03-15 2015-09-02 A network of smart machines
HK16106098.0A HK1218174A1 (zh) 2013-03-15 2016-05-30 智能機器的網絡
JP2019053265A JP6790160B2 (ja) 2013-03-15 2019-03-20 インテリジェントマシンのネットワーク
US16/741,551 US11544533B2 (en) 2013-03-15 2020-01-13 Network of intelligent machines
US18/091,654 US20230153657A1 (en) 2013-03-15 2022-12-30 Network of intelligent machines

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
US13/843,784 US10534995B2 (en) 2013-03-15 2013-03-15 Network of intelligent machines

Related Child Applications (1)

Application Number Title Priority Date Filing Date
US16/741,551 Continuation US11544533B2 (en) 2013-03-15 2020-01-13 Network of intelligent machines

Publications (2)

Publication Number Publication Date
US20140279717A1 US20140279717A1 (en) 2014-09-18
US10534995B2 true US10534995B2 (en) 2020-01-14

Family

ID=50336525

Family Applications (2)

Application Number Title Priority Date Filing Date
US13/843,784 Active 2034-05-06 US10534995B2 (en) 2013-03-15 2013-03-15 Network of intelligent machines
US16/741,551 Active 2034-05-14 US11544533B2 (en) 2013-03-15 2020-01-13 Network of intelligent machines

Family Applications After (1)

Application Number Title Priority Date Filing Date
US16/741,551 Active 2034-05-14 US11544533B2 (en) 2013-03-15 2020-01-13 Network of intelligent machines

Country Status (10)

Country Link
US (2) US10534995B2 (ru)
EP (2) EP4044066A1 (ru)
JP (2) JP6528040B2 (ru)
CN (1) CN105122277B (ru)
BR (1) BR112015023372B1 (ru)
CA (1) CA2903041C (ru)
HK (1) HK1218174A1 (ru)
IL (1) IL241056B (ru)
RU (1) RU2669527C2 (ru)
WO (1) WO2014149510A2 (ru)

Families Citing this family (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US10217013B2 (en) * 2014-06-30 2019-02-26 Carnegie Mellon University Methods and system for detecting curved fruit with flash and camera and automated image analysis with invariance to scale and partial occlusions
US20160110657A1 (en) * 2014-10-14 2016-04-21 Skytree, Inc. Configurable Machine Learning Method Selection and Parameter Optimization System and Method
JP6662902B2 (ja) * 2015-06-05 2020-03-11 グーグル エルエルシー 空間的変換モジュール
ITUB20155003A1 (it) * 2015-10-29 2017-04-29 Ser Mac S R L Apparato per il rilevamento di prodotti ortofrutticoli; in particolare di prodotti ortofrutticoli difettosi.
CN106650948A (zh) * 2016-12-09 2017-05-10 曙光信息产业(北京)有限公司 一种机器学习中避免大数据冗余的方法
CN110235137A (zh) * 2017-02-24 2019-09-13 欧姆龙株式会社 学习数据获取设备和方法、程序和存储介质
US10543942B2 (en) * 2017-11-21 2020-01-28 Fulfil Solutions, Inc. Product handling and packaging system
US11574235B2 (en) 2018-09-19 2023-02-07 Servicenow, Inc. Machine learning worker node architecture

Citations (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5845002A (en) 1994-11-03 1998-12-01 Sunkist Growers, Inc. Method and apparatus for detecting surface features of translucent objects
JP2003316746A (ja) 2002-04-25 2003-11-07 Toshiba Corp 人物認識システム
RU72084U1 (ru) 2007-12-03 2008-03-27 Илья Самуилович Кабак Доменная нейронная сеть
US20100002604A1 (en) * 1999-10-28 2010-01-07 Lightwaves Systems, Inc. Method for routing data packets using an ip address based on geo position
US7739207B2 (en) 2006-07-11 2010-06-15 International Business Machines Corporation Network autonomous learning system that allows computers to share learned knowledge such that errors and potential problems are identified and resolved
US20100315207A1 (en) 2009-06-15 2010-12-16 Qualcomm Incorporated Sensor network management
US20110179052A1 (en) 2010-01-15 2011-07-21 Canon Kabushiki Kaisha Pattern identification apparatus and control method thereof
US20120197823A1 (en) 2004-12-10 2012-08-02 Numenta, Inc. Methods, architecture, and apparatus for implementing machine intelligence and hierarchical memory systems
US20120330885A1 (en) 2006-02-10 2012-12-27 Numenta, Inc. Sequence learning in a hierarchical temporal memory based system
US20140033819A1 (en) * 2008-02-01 2014-02-06 Kraft Foods R & D, Inc. Apparatus For Determining The Texture Of Food Material

Family Cites Families (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP4411900B2 (ja) * 2003-08-08 2010-02-10 ソニー株式会社 電子機器間の相互成長システム、電子機器及びロボット装置
JP4544847B2 (ja) * 2003-11-20 2010-09-15 ソニー・エリクソン・モバイルコミュニケーションズ株式会社 電子機器及びシステム
US7656810B2 (en) * 2005-03-25 2010-02-02 Microsoft Corporation System and method for monitoring and reacting to peer-to-peer network metrics
US8428310B2 (en) * 2008-02-28 2013-04-23 Adt Services Gmbh Pattern classification system and method for collective learning
CN101973031B (zh) * 2010-08-24 2013-07-24 中国科学院深圳先进技术研究院 云机器人系统及实现方法
CN102571859A (zh) * 2010-12-29 2012-07-11 鸿富锦精密工业（深圳）有限公司 通过云计算控制机器人的系统及方法
JP5923723B2 (ja) * 2011-06-02 2016-05-25 パナソニックＩｐマネジメント株式会社 人物属性推定システム、人物属性推定装置、及び人物属性推定方法

Patent Citations (12)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5845002A (en) 1994-11-03 1998-12-01 Sunkist Growers, Inc. Method and apparatus for detecting surface features of translucent objects
US20100002604A1 (en) * 1999-10-28 2010-01-07 Lightwaves Systems, Inc. Method for routing data packets using an ip address based on geo position
JP2003316746A (ja) 2002-04-25 2003-11-07 Toshiba Corp 人物認識システム
US20120197823A1 (en) 2004-12-10 2012-08-02 Numenta, Inc. Methods, architecture, and apparatus for implementing machine intelligence and hierarchical memory systems
US20120330885A1 (en) 2006-02-10 2012-12-27 Numenta, Inc. Sequence learning in a hierarchical temporal memory based system
US7739207B2 (en) 2006-07-11 2010-06-15 International Business Machines Corporation Network autonomous learning system that allows computers to share learned knowledge such that errors and potential problems are identified and resolved
RU72084U1 (ru) 2007-12-03 2008-03-27 Илья Самуилович Кабак Доменная нейронная сеть
US20140033819A1 (en) * 2008-02-01 2014-02-06 Kraft Foods R & D, Inc. Apparatus For Determining The Texture Of Food Material
US20100315207A1 (en) 2009-06-15 2010-12-16 Qualcomm Incorporated Sensor network management
CN102461071A (zh) 2009-06-15 2012-05-16 高通股份有限公司 用于配置网络中的传感器的方法
US20110179052A1 (en) 2010-01-15 2011-07-21 Canon Kabushiki Kaisha Pattern identification apparatus and control method thereof
JP2011145958A (ja) 2010-01-15 2011-07-28 Canon Inc パターン識別装置及びその制御方法

Non-Patent Citations (12)

* Cited by examiner, † Cited by third party
Title
Bugra Gedik, et al., "ASAP: An Adaptive Samploing Approach to Data Collection in Sensor Networks", IEEE Transactions on Parallel and Distributed Systems, vol. 18, No. 12, Dec. 2007, pp. 1766-1783.
Carpenter, Gail A., and Stephen Grossberg. "A massively parallel architecture for a self-organizing neural pattern recognition machine." Computer vision, graphics, and image processing 37.1 (1987): 54-115. *
Dean, Jeffrey, et al. "Large scale distributed deep networks." Advances in Neural Information Processing Systems. 2012. *
Eduardo F. Nakamura, et al., "Information Fusion for Wireless Sensor Networks: Methods, Models, and Classifications", ACM Computing Surveys, Vo. 39, No. 3, Article 9, Aug. 2007, 55 pages.
Hannah, et al., "Decisions in Condition Monitoring-An Examplar for Data Fusion Architecture", Manchester School of Engineering University of Manchester, ISIF 2000, 8 pages.
Hannah, et al., "Decisions in Condition Monitoring—An Examplar for Data Fusion Architecture", Manchester School of Engineering University of Manchester, ISIF 2000, 8 pages.
Huang, Cheng-Lung, and Jian-Fan Dun. "A distributed PSO-SVM hybrid system with feature selection and parameter optimization." Applied Soft Computing 8.4 (2008): 1381-1391. *
International Search Report, corresponding to PCT/US2014/019134, dated Oct. 28, 2014, 4 pages.
Nicola Bicocchi,et al., "Pervasive Self-Learning with Multi-Modal Distributed Sensors", 2008 Second IEEE International Conference on Self-Adaptive and Self-Organizing Systems Workshops, 6 pages.
Search Report corresponding to RU2015139090, dated Feb. 15, 2018, 2 pages.
Valera, Maria, and Sergio A. Velastin. "Intelligent distributed surveillance systems: a review." IEE Proceedings-Vision, Image and Signal Processing 152.2 (2005): 192-204. *
Valera, Maria, and Sergio A. Velastin. "Intelligent distributed surveillance systems: a review." IEE Proceedings—Vision, Image and Signal Processing 152.2 (2005): 192-204. *

Also Published As

Publication number Publication date
CN105122277A (zh) 2015-12-02
EP4044066A1 (en) 2022-08-17
BR112015023372A2 (pt) 2017-07-18
WO2014149510A2 (en) 2014-09-25
HK1218174A1 (zh) 2017-02-03
IL241056B (en) 2020-09-30
BR112015023372B1 (pt) 2022-08-16
WO2014149510A3 (en) 2014-12-24
RU2669527C2 (ru) 2018-10-11
EP2973239A2 (en) 2016-01-20
CA2903041C (en) 2022-05-03
US20140279717A1 (en) 2014-09-18
US20200226448A1 (en) 2020-07-16
CA2903041A1 (en) 2014-09-25
IL241056A0 (en) 2015-11-30
RU2015139090A3 (ru) 2018-03-12
JP2016515261A (ja) 2016-05-26
US11544533B2 (en) 2023-01-03
CN105122277B (zh) 2019-03-22
JP2019149173A (ja) 2019-09-05
JP6790160B2 (ja) 2020-11-25
JP6528040B2 (ja) 2019-06-12
RU2015139090A (ru) 2017-04-19

Similar Documents

Publication Publication Date Title
US11544533B2 (en) Network of intelligent machines
Dubey et al. Application of image processing in fruit and vegetable analysis: a review
Kodagali et al. Computer vision and image analysis based techniques for automatic characterization of fruits-a review
Khojastehnazhand et al. Development of a lemon sorting system based on color and size
US7012256B1 (en) Computer assisted bag screening system
Wang et al. A multimodal machine vision system for quality inspection of onions
CN109886555A (zh) 食品安全的监测方法及装置
Kutyrev et al. Recognition and Classification Apple Fruits Based on a Convolutional Neural Network Model.
Sun et al. [Retracted] Research on Machine Vision System Design Based on Deep Learning Neural Network
Mariyam et al. Automatized Food Quality Detection and Processing System Using Neural Networks
Gan et al. An automated cucumber inspection system based on neural network
Kanali et al. Three-dimensional shape recognition using a charge-simulation method to process primary image features
Lopez-Juarez et al. A hybrid non-invasive method for internal/external quality assessment of potatoes
US20230153657A1 (en) Network of intelligent machines
JP7034529B1 (ja) 学習モデルの生成方法、学習モデル、検査装置、検査方法およびコンピュータプログラム
Buyukarikan et al. Classification of physiological disorders in apples using deep convolutional neural network under different lighting conditions
Li Development of a structured illumination reflectance imaging system for enhanced detection of subsurface and surface defects in apple fruit
Caguioa et al. Duck Egg Quality Classification Based on its Shell Visual Property Through Transfer Learning Using ResNet-50
Feldman et al. A scalable, low‐cost phenotyping strategy to assess tuber size, shape, and the colorimetric features of tuber skin and flesh in potato breeding populations
Jagadeesan et al. Perishable Food Monitoring Model Based On Iot And Deep Learning To Improve Food Hygiene And Safety Management
Begum et al. Deep learning based image processing solutions in food engineering: A review
Kanali et al. Shape identification using a charge simulation retina model
Villaseñor-Aguilar et al. Fuzzy Classification of Color Carrots (Dacus Carota) using Raspberry Pi towards Farming 4.0
Paliwal Digital image analysis of grain samples for potential use in grain cleaning
Pla¹ et al. A machine vision system for on-line fruit colour classification

Legal Events

Date Code Title Description
AS Assignment

Owner name: QYLUR INTELLIGENT SYSTEMS, INC., CALIFORNIA

Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNORS:SAGI-DOLEV, ALYSIA M.;ZWEIG, ALON;SIGNING DATES FROM 20180306 TO 20180307;REEL/FRAME:045132/0263

STPP Information on status: patent application and granting procedure in general

Free format text: DOCKETED NEW CASE - READY FOR EXAMINATION

STPP Information on status: patent application and granting procedure in general

Free format text: NOTICE OF ALLOWANCE MAILED -- APPLICATION RECEIVED IN OFFICE OF PUBLICATIONS

STPP Information on status: patent application and granting procedure in general

Free format text: PUBLICATIONS -- ISSUE FEE PAYMENT VERIFIED

STCF Information on status: patent grant

Free format text: PATENTED CASE

MAFP Maintenance fee payment

Free format text: PAYMENT OF MAINTENANCE FEE, 4TH YR, SMALL ENTITY (ORIGINAL EVENT CODE: M2551); ENTITY STATUS OF PATENT OWNER: SMALL ENTITY

Year of fee payment: 4